On extracting commuter information from GPS motion data

Commuters rely on realistic and real-time information in order to optimize the time spent on commuting between home and work. Delays in (urban) transport and congestion for individual motorized transport are a major issue for unnecessary long travel times. While some of these delays occur randomly, there is also a systematic component. In this paper we describe a data-driven approach to analyze positions of an individual collected using GPS to obtain information on the individual’s typical routes, typical schedules and the used mode of transport. Furthermore, we propose an approach to model the probability of an event like missing a train as a function of time. This allows to optimize the expected commuting time based solely on the commuters motion history. Suitability of the approach is demonstrated in a real world application based on a dataset comprising six weeks of GPS tracks.